This paper describes the creation of DUO_DE A1, a corpus based on A1-level learning material from the Deutsch-Uni Online (DUO) language courses for German as a foreign language. We split the material into small segments and manually annotated each with fine-grained information such as the type of segment (e.g. task description, description of grammar), the medium (e.g. text, table, audio), the text units it contains (e.g. words, phrases, sentences) and other special features (e.g. marking cloze texts). Furthermore, we automatically tokenized, POS tagged and lemmatized the corpus and compared the performance of three models on these steps for different kinds of segments. We publish the created corpus in a manner that respects copyright, releasing all structural features, metadata and POS tags.

We present an approach to the automated scoring of a German Written Elicited Imitation Test, designed to assess literacy-dependent procedural knowledge in German as a foreign language. In this test, sentences are briefly displayed on a screen and, after a short pause, test-takers are asked to reproduce the sentence in writing as accurately as possible. Responses are rated on a 5-point ordinal scale, with grammatical errors typically penalized more heavily than lexical deviations. We compare a rule-based model that implements the categories of the scoring rubric through hand-crafted rules, and a deep learning model trained on pairs of stimulus sentences and written responses. Both models achieve promising performance with quadratically weighted kappa (QWK) values around .87. However, their strengths differ – the rule-based model performs better on previously unseen stimulus sentences and at the extremes of the rating scale, while the deep learning model shows advantages in scoring mid-range responses, for which explicit rules are harder to define.